New Home Sales Up (But Beware Double Digit Monthly Gains)

Commerce Department reported that April new-home sales jumped an unexpected 16.2% -- the biggest monthly gain in 14 years. The year over year drop in sales change was a drop of 10.6%.

By no coincidence, the median home price dropped a 11.1% from the prior month -- the largest amount on record. The big price decline implies that builders are slashing prices to move the huge overhang of unsold inventory. Inventory of unsold new homes fell to 6.5 months (seasonally adjusted) from 8.1.

Regional sales were even more telling. As we noted earlier in the week, the greater NY region (and NYC Condo/Coop sales) are skewing the national data. Year-over-year sales in the Northeast surged 43.1%; sales were down 28.1% in the Midwest, and 25.4% in the West. Sales fell marginally (3.4%) in the South.

~~~

In the past, we have cautioned that double digit gains of new home sales are VERY unreliable monthly data points. (It is mostly due to the way the Builders self report their sales to Commerce).

Have a look at this analysis we did a few years ago:  It turns out that whenever New Home
Sales jump double digits, it usually reflects a mean reversion from the prior (or subsequent) month's reportage. Indeed, over the past 15 years of data, we found that a mean
regression followed nearly every double digit monthly gains. Typically, the
subsequent month's data was significantly lowered -- flat to negative
in nearly every case:

>
New Homes Sales

onth, Year	Double Digit Gain	Subsequent Month	Increase / Decrease
June 2003	10.7%	July 2003	(-2.1%)
December 2000	11.7%	January 2001	(-4.8%)
July 2000	11.9%	August 2000	(-4.4%)
November 1998	11.4%	December 1998	(-4.6%)
January 1998	10%	February 1998	(-0.7%)
March 1995	10.2%	April 1995	0.8%.
*February 1994;	10.82%	March 1994	8.89%
April 1993	16.45%	May 1993	(-10.70%)
September 1993;	12.56%	October 1993	(-3.03%)
January 1992	21.15%	February 1992	(-5.47%)

*February 1994 was the one exception -- it was followed by strong
March and April data -- but came after January 1994, which has the
honor of being the very worst month ever in the history of the Census Construction data: Down -23.77%.

In other words, stick with the two or three month
average.

>

>

Sources:
New Residential Sales
Commerce Department Census Division
http://www.census.gov/newhomesales
http://www.census.gov/const/newressales.pdf

U.S. April new-home sales unexpectedly jump 16%
Rex Nutting
MarketWatch, 10:00 AM ET May 24, 2007
http://tinyurl.com/36dt2d

Home Sales Soar by Record Amount
Thursday May 24, 10:09 am ET
By Terence Hunt, AP White House Correspondent
http://biz.yahoo.com/ap/070524/economy.html?.v=17

Comments

[Johnny Bitc...:]

Good Info.

Regarding the data:

1. If there is self-reporting bias, it is not just the double digit months data that are unreliable, it is all months.

2. What you are calling mean reversion isn't really mean reversion if it is caused by self-reporting bias - its just unreliable data, though it may be parading as mean reversion if the unreliability is systematically biased and then systematically biased in the reverse direction the following month.

Do you have a particular theory that the builders are gaming the BLS report and producing numbers that they wish to see for whatever reason ?

If so, and they are willing to play games by moving sales in to the current month or out to next month without regard to the actual sales transaction date, might they not just lie outright ?

Are there any checks on blatant lying on these reports or are they confidential ? It would be interesting to know what the legal arrangement for the information exchange is, and whether it falls under Sarbanes-Oxley rules, etc.

Just wondering if you know any source that may have reconciled this data with other available sources.

John.

2007 May 24 01:19 PM

[Johnny Bitc...:]

Hey Barry, if you want to add value (and I know you do !), you could get the numbers for the typical delta (or at least the direction of the delta) for each of the month-month pairings above.

If it's near zero or even positive, you'd have some pretty good evidence of bias (self-reporting or otherwise) and maybe at some point your "BLS data is phony" jihad will get more traction.

One of the problems with you being the leader of this jihad is that your own methods are often sloppy, incomplete, spurious, phony, emotional, biased, anchored, and sometimes downright foolish. This element of hypocrisy works against you in this movement.

Just as an example, if you are going to go to the trouble to identify these month over month peculiarities, why don't you follow through and provide the typical increase/decrease for the month pairings ?

Surely this data is not hard to obtain.

Surely you understand that your just slinging b.s. unless you provide all the relevant figures to determine whether your implied accusations have any basis in statistical reality.

Drawing conclusions based on statistics is difficult enough without having to wonder whether they are incomplete and hand selected to promote the arguments of the presenter.

I challenge you to provide the missing statistics: the normal month-month comparisons as an additional column in the table above. Nothing fancy......you don't have to provide t-stats or p-values, just the normal delta for those months.

Easy enough, right ?

No disrespect, but if you don't provide that simple information, I will just assume you are a phony, chit-chatty, market watching fool......rather than a serious professional.

Cheers.

John.

2007 May 24 03:50 PM

[BillCust:]

Guys, here is the relevant part of the report, paste it to XLS if it's hard to read here. Basically shows the entire surprise gain was in the South region (if you put in Flat 440k relative to March, you get the 860k forecast by economist). Moreover, the price is very low relative to the rest of the country (

2007 May 24 04:09 PM